Description Templates for Common Graphics: Final Report
Editor’s Note: This is a summary of completed DIAGRAM subcontract work, submitted by Steve Landau. Find the full report and links to the surveys for the templates by visiting the DIAGRAM Development web page.
Research Team: Steve Landau, Principal Investigator, with Lucia Hasty, Yue-Ting Siu, Josh Miele, and Val Morash
Rather than being expected to develop unique prose for each image they are tasked to describe, image describers should be supported to describe according to established guidelines. This will simplify their work, and may reduce cognitive load for readers. This project has created software that prompts describers with image description templates for common types of graphics, based on the NCAM STEM description guidelines. Image describers respond to questions regarding the basic characteristics of a diagram, and are guided through a fill-in-the-blanks process, which generates the first paragraph of an image description. The describer is then free to edit that paragraph and to append additional explanatory language. An assessment program built into the proposed project provides feedback on the consistency and accuracy of descriptions provided by users of the templates.
In analyzing the data generated during this study, the researchers determined that guided and unguided descriptions were equally likely to contain syntactic or content errors, contain text or table elements, and provide a description of the data and data trends. Guided and unguided descriptions were also equal in how long they took to create and their text word counts. However, guided descriptions were more likely to contain important image information, including the chart type, title, caption, and units. Furthermore, the participants found description using guidance/templates to be significantly easier, and they preferred this method over unguided description.
The current report assesses the production of image descriptions through unguided and guided interfaces, and compares the resulting descriptions on their accuracy and completeness. The guided descriptions were no more or less accurate than the unguided descriptions, but were more complete. The current research does not address whether unguided or guided descriptions promoted comprehension and usability. Further research will be needed to investigate what properties of descriptions work best for students and teachers, and whether these properties are encouraged or discouraged by our guided-description system.
For further information, please contact <info [at] touchgraphics [dot] com>.